I’ve installed the CUDA toolkit via anaconda, but it keeps saying nvcc isn’t recognised as a command, anyone know what else I need to install or configure?
the PATH variable
you’ll have to figure out where anaconda put nvcc, and then configure your path to point to that directory
Any idea where the path is that I need to configure and roughly how it should look like?
are you on linux or windows?
Windows
use the windows file find utility (i.e. the search bar in the upper right hand corner of the file explorer) to search your hard drive for nvcc.exe
Eh, I keep getting various errors and I have no idea what I’m doing anymore, I think I’ll give up on this for now.
I’ll try and eleborate my situation in more detail in the hopes of maybe there being an obvious solution:
I’m just trying to get this video/image interpolation thing to work:
But the creator seems to have not used Windows for a long while and has done his work with Ubuntu in mind. Do you happen to have any idea how I could set it up on my windows 7 machine or would it be best to just install ubuntu and play with that?
The Anaconda cudatoolkit is just the Python-CUDA bindings for Numba. You would need to install Visual Studio and the NVIDIA CUDA Toolkit. I recommend using Ubuntu (I’ve done performance testing on an image processing routine that ultimately uses the scikit-image warp function - it runs twice as fast on Linux). After you get your NVIDIA drivers working, install the “deb (local)” package and use the meta-package “cuda-toolkit-10-0” so it doesn’t overwrite your drivers. Keep in mind that the Anaconda cudatoolkit is still at version 9.2, not sure if it will work with NVIDIA CUDA 10.